Neural inverse-based fast control of quadrotors with actuator saturation

Abstract

This study investigates the fast attitude control for quadrotors subject to possible actuator saturations. A low-complexity neural dynamic inverse is constructed to decouple attitude dynamics without applying the conventional small-angle assumption. Aiming at decoupled dynamics with mismatched disturbances, an improved terminal sliding mode controller is designed to realize the fast attitude regulation. In particular, a novel anti-windup mechanism and a finite-time disturbance observer are constructed, respectively, to evaluate motor saturations and estimate mismatched disturbances online, whose information is employed to optimize the baseline terminal sliding mode controller. The effectiveness of the proposed control has been theoretically proven and practically tested.